Multiple Timeframes

[Hlm 10]

  jasont said:

I am actually disappointed I didn't read about Hlm's intentions of creating a thread on the topic himself prior to posting this thread. I believe Hlm can do a better job than myself in explaining market his application of Market Fractals. In fact the discussion you two had in the thread is probably more along the lines of what I hoped might have developed here.

 

PS. I am more than happy to let this one die if Hlm wants to create another one.

I see no need to stop using this one at the moment. If a topic comes up that starts to change direction we can just split it out into another thread. I do agree, however, we should try not to get caught up on definitions since depending on your prior background two people can be looking at the same thing and argue while in reality they are in perfect agreement. Let's keep these minds open, you never know what you may find. I am still on vacation but will try to get something up within the next couple of days. However, let's keep this thread moving along in the mean time. I don't want this thread to turn into "what does Hlm say". That can cut off the open mindedness and sharing at the knees.

[BlowFish 10]

I like CandleWispers description (note description rather than definition). However you sample market data you find similar patterns of higher highs lower lows. You see congestions on a 5 minute chart that look quite similar to a congestion on a weekly chart.

 

Another way of looking at things is does a yearly distributions of market data resemble monthly weekly or daily distributions? (to me they appear to). Mandlebrot (who I guess is the father of the fractal) posited that financial data (he researched cotton) follows a "Lévy skew alpha-stable distribution with α equal to 1.7". I understand that would make it self similar.

 

Personally I think of markets as being 'fractal', not in any strict mathematical sense, but simply as you see similar phenomena regardless of how you sample the data (which is a way of saying on all time frames).

[CandleWhisperer 10]

  DbPhoenix said:

Well, I don't see the market as being fractal since it's missing the self-similarity element. And I disagree that there is any such thing as "noise". But, as I said, all this may be entirely off-topic.

 

There is no such thing as "noise". If you believe that price is continuous and every price is valid because a traded happened there. People make price incongruent by using various arbitrary timeframes. This is because unlike price, man is a temporal being. But if these timeframes are all equally arbitrary they need to have similar properties, otherwise one would be more or less arbitrary than another. If all must exhibit similar properties, that means if you look at a 5 minute chart minus the scales it has to look like a 23 minute chart minus the scales.

 

The result is similar to looking at a coastline from space and then looking at from a small child. The child sees more detailed coastline that resembles the one seen from space. The Ant on the coastline sees an even more detailed coastline that resembles the one seen by a child. That resembles the coastline as seen from the humble. This is the definition of self similarity. It is also a necessary condition of a fractal.

 

We have established that any X, where X=timeframe must be similar to any x+-i, where i= any number plus or minus, if all are to be equally arbitrary. This must lead to self similarity. So we can conclude that markets are fractal.

[DbPhoenix 42]

  CandleWhisperer said:

There is no such thing as "noise". If you believe that price is continuous and every price is valid because a traded happened there. People make price incongruent by using various arbitrary timeframes. This is because unlike price, man is a temporal being. But if these timeframes are all equally arbitrary they need to have similar properties, otherwise one would be more or less arbitrary than another. If all must exhibit similar properties, that means if you look at a 5 minute chart minus the scales it has to look like a 23 minute chart minus the scales.

 

The result is similar to looking at a coastline from space and then looking at from a small child. The child sees more detailed coastline that resembles the one seen from space. The Ant on the coastline sees an even more detailed coastline that resembles the one seen by a child. That resembles the coastline as seen from the humble. This is the definition of self similarity. It is also a necessary condition of a fractal.

 

We have established that any X, where X=timeframe must be similar to any x+-i, where i= any number plus or minus, if all are to be equally arbitrary. This must lead to self similarity. So we can conclude that markets are fractal.

 

If only price were a coastline.....

[Hlm 10]

I had a little time this afternoon to throw this together.

I am purposely keeping this over simplified and without specifics.

I want to see what discussion naturally comes from this.

I purposely don't plan on replying until more discussion has taken place.

 

[WRR 10]

Here are a few things I look at on multiple timeframes.

 

First of all, I sometimes use a fractal indicator that shows fractal high and fractal low points. The indicator draws an up arrow on top of local fractal high bars (a bar which has at least two bars with lower highs to the left of it and at least two bars with lower highs to the right of it). Similarly it draws a down arrow on bottom of local fractal low bars (a bar which has at least two bars with higher lows to the left of it and at least two bars with higher lows to the right of it). I say sometimes I use this indicator because I usually can quickly spot fractal highs and lows without using it.

 

Now I will use these points as local support and resistance and also as points to draw fibs. So I might draw SR and fib off 1hr chart then do the same off 4hr and/or daily chart. Sometimes the fractal high and low points will be the same in different timeframes, giving them more weight, but often they will be different.

 

This now shows me multiple possible targets for stop losses, breakouts or retracements with greater weight going to the nearby SR and fibs on the higher timeframe charts.

 

It gives you a big picture view before you pull the trigger.

 

As for the self-similarity idea, Elliott Wave is a good example, In wave one on a larger time frame chart, one could have a complete 5 wave pattern on a smaller timeframe chart.

 

A further use for multiple timeframe charts is as a guide for trade direction confirmation. If the current bar on all the timeframes you use are in the same direction, then the trade has stronger momentum and directional bias.

 

WRR

[BlowFish 10]

  DbPhoenix said:

If only price were a coastline.....

 

You know if you plot every tick of the S&P on a 1"x1" scale and look at it from space it looks just like Norways coastline Your comment gave me my first smile of the day....thanks.

[jasont 10]

Ok guys this is exactly where I was hoping the thread would head and its great to see others putting in their ways to look at multiple time frames. To keep it on track I think it is best we avoid the use of the word fractal as some people have different definitions of the word. At the same time I think we are all talking about the same thing so it might be best to refer to things as multiple time frames. I am all for a different thread discussing the difference in meanings of the word fractal if that is where people want to discuss it.

 

Great post Hlm as it really hits on the core principle behind looking at the market through multiple time frames. I actually want to go a bit in depth on Hlm's question on how does one define the point he mentioned.

 

I will start by saying that I use the largest time frame for broad trend and the middle time frame to set up the market. I then use the smallest time frame to trigger the entry. I don't want this to become a he entered here or there thread because there are stacks of those threads already.

 

I will take some action from yesterdays market as an example of how I see the actual time frames together. Many traders only take the extreme points on a chart as the support and resistance levels.

 

By that lets use the example I have made below. The top chart is a 5 minute chart and the bottom chart is a 1 minute chart. A low is made at Low 2 on both charts. Many traders would only take the low of that point as the support area. However if one looks at the 5 minute chart it looks like a simple bounce, which it is. However when looking at point 2 on the 1 minute chart we can see a range has been built.

 

Now using the traditional trading idea that markets break out of ranges and then retest them, we can see that the market didn't immediately test the range it made in the 1 minute chart. The market proceeds to make a higher high encouraging one, with the context of the current market, to think that the market is moving higher for the day.

 

Now many traders would only look for the market to reach the previous high and bounce from that, which is a valid idea, and the next level of support is the previous low. However looking at the 1 minute chart we see a range at that low. Typically on tests of ranges, we wouldn't want to see the market reach into a range too far otherwise it makes the pullback trade less likely.

 

So at Low 3, instead of looking for the market to reach the extreme, one can look at that low as a range on the 1 minute chart. So when we reach that Low 3, we can treat it as a pullback after a range breakout.

 

Now for me, this occurs on all time frames. Most traders already know the market ranges, moves up, ranges, moves down. So looking for these ranges on multiple time frames helps one find the ideal points to watch in the market.

 

That is personally how I define the point that Hlm asks the question of. I am sure there are numerous other ways to use multiple time frames so hopefully this gets some more posts in.

[OAC 10]

  WRR said:

Here are a few things I look at on multiple timeframes.

 

First of all, I sometimes use a fractal indicator that shows fractal high and fractal low points. The indicator draws an up arrow on top of local fractal high bars (a bar which has at least two bars with lower highs to the left of it and at least two bars with lower highs to the right of it). Similarly it draws a down arrow on bottom of local fractal low bars (a bar which has at least two bars with higher lows to the left of it and at least two bars with higher lows to the right of it). I say sometimes I use this indicator because I usually can quickly spot fractal highs and lows without using it.

 

In order to keep this discussion productive, we need to keep our definitions consistent, so far we have only touched upon the fractal nature of the market as self-similarity across different time-frames.

Your usage of the terms of fractal high, fractal low, and the associated fractal indicator, I believe.are terms popularize by Bill Williams, author of "Trading Chaos". These can be easily be called pivot high, pivot low ..etc, unless you are a vendor who is trying to sell more subscriptions. :rofl:

I think Dbphoenix was correct in saying the term "fractal" was often being misused.

[firewalker 10]

  CandleWhisperer said:

The result is similar to looking at a coastline from space and then looking at from a small child. The child sees more detailed coastline that resembles the one seen from space. The Ant on the coastline sees an even more detailed coastline that resembles the one seen by a child. That resembles the coastline as seen from the humble. This is the definition of self similarity. It is also a necessary condition of a fractal.

 

According to the Wikipedia definition of a fractal, self-similarity is indeed a necessary condition, which means that the fractal can be split into "parts, each of which are a reduced-size copy of the whole."

 

And that's where I think the misinterpretation takes place, because although the market may exhibit similar patterns along different time frames, the element of recursion does not take place.

 

See the home-made chart

Just my 2c.

 

[CandleWhisperer 10]

  firewalker said:

According to the Wikipedia definition of a fractal, self-similarity is indeed a necessary condition, which means that the fractal can be split into "parts, each of which are a reduced-size copy of the whole."

 

And that's where I think the misinterpretation takes place, because although the market may exhibit similar patterns along different time frames, the element of recursion does not take place.

 

See the home-made chart

Just my 2c.

 

 

Nice chart. But think about Elliot Wave.

 

A wave 1 on a 15 minute chart can be seen to be an entire 5 wave pattern on the smaller 5 minute chart. Of course the argument then shifts to the existence of Elliot Waves in the first place.

[firewalker 10]

  CandleWhisperer said:

A wave 1 on a 15 minute chart can be seen to be an entire 5 wave pattern on the smaller 5 minute chart. Of course the argument then shifts to the existence of Elliot Waves in the first place.

 

Well there's no point in discussing something hypothetical when there a plenty of charts out there each day

[atto 20]

  CandleWhisperer said:

Nice chart. But think about Elliot Wave.

 

A wave 1 on a 15 minute chart can be seen to be an entire 5 wave pattern on the smaller 5 minute chart. Of course the argument then shifts to the existence of Elliot Waves in the first place.

That would be incorrect. Elliott Wave doesn't consist of just the classic 5 wave pattern, but a myriad of other patterns to accommodate for the fact that markets do not move in 3 wave impulses (with pullbacks in between). In fact, "good" EW charts are usually extremely complex. The 5 wave pattern does happen, but it's not the norm at all (not enough to use it as the basis of a fractal argument).

[Hlm 10]

To keep this thread moving smoothly...

 

For all the purist that can't get past the mathematical definition...how about the market consists of extreme fractal influences that often roughly resemble self-similarity. An individual can define "often" and "roughly" on their own and preferably not in this thread. As for the sake of this thread moving smoothly, let's stick to the words up trend, down trend, and time frames.

 

Now let us continue.

Edited January 6, 2009 by Hlm

[WRR 10]

  OAC said:

In order to keep this discussion productive, we need to keep our definitions consistent, so far we have only touched upon the fractal nature of the market as self-similarity across different time-frames.

Your usage of the terms of fractal high, fractal low, and the associated fractal indicator, I believe.are terms popularize by Bill Williams, author of "Trading Chaos". These can be easily be called pivot high, pivot low ..etc, unless you are a vendor who is trying to sell more subscriptions. :rofl:

I think Dbphoenix was correct in saying the term "fractal" was often being misused.

 

Hi OAC,

I am not a vendor. I was just explaining how I use multiple timeframe charts. I agree with you that If I had used the term pivot high and pivot low instead, it would make no difference to the method I explained. It is a way I use to identify high and low points for support and resistance and to draw fibs. And this often results in different levels for the same time period depending on the chart time frame.

 

Cheers,

WRR

[OAC 10]

We should be grateful that the market is fractal in nature and not fractal in the truest sense. If it were, we would all be out of work.

[Head2k 10]

I am a beginner and during my learning I was also thinking of market "fractality". I had no problems with recognising the existence of different price positions on diferent time frames, but my problem was to make order in it, that means to decide what the hell I want to trade then, and how to trade it.

Anyway, as a remark to the first pages of this thread, I think definitions are very important. Definitions show the way of your perception. Therefore I think a lot about definitions and terms I adopt in my trading, or learning process.

Now I don't think of different time frames, I think of different scales. When I think of "scale" instead of "time frame" it is obvious that time per se is not that relevant. It is also obvious that interval of a particular chart doesn't matter. What matters more is the scale of the action.

As for reversal being breakout, it is also a matter of preception. One can use a breakout of some sort to time entry into a reversal, but he must keep in mind that it is the reversal that he is trading, not the breakout. He must keep in mind the scale that he trades.

For me the result of thinking of diferent time frames was that I would think of scales instead and that I will restrict my trading significantly. I put a down limit on scale I am willing to trade and I decided to plan all my potential trades before open. Not general rules but concrete trades.

Not sure if it fits in this thread but that's my (beginner's) take on this topic.

 

EDIT:

Also when thinking of diferent time frames, I think the action is self similar only to certain degree. Larger time frames smooth the action (or traders decisions), and to see some sorts of behavior one has to watch the action in sufficient detail.

Edited January 6, 2009 by Head2k

[Hlm 10]

Very good point Head2k. Time Frame implies that it's sorted by time which I would disagree with. However, many times it's the easiest way to get someone to understand the topic. For me personally, I don't put any time/points/etc label on it when trading. It always starts off with a trend as previously defined. At which point I find the trends above and below to define my trade. Instead of trying to push arbitrary settings on the market I let it define itself. If I ever label a time frame to share with others it is usually based on a time/range where the trend can be seen clearly while not adding too much detail of the smaller trends. So when it comes down to it...for me at least, it makes more sense to use the phrase "multi-trends" to describe this concept. Of course it could be the white wine speaking.

:cheers:

[BlowFish 10]

I think people are getting confused about what a fractal actually is. You can iterate an equations with complex components (having a real part and an imaginary part) and some of the resultant sets have edges that are fractal. However at the time they where discovered the term fractal had not been coined. These are mathematical fractals.

 

http://en.wikipedia.org/wiki/Mandelbrot_set

 

Mandlebrot noticed from other research on datasets from applied fields (rather than mathematics) that many (including financial data series) had the same characteristics of self similarity and 'fat tails' in their distributions. In 1975, Mandelbrot coined the term fractal to describe these structures, and published his ideas in Les objets fractals, forme, hasard et dimension (1975; an English translation Fractals: Form, Chance and Dimension was published in 1977). From http://en.wikipedia.org/wiki/Benoît_Mandelbrot (see 'later years').

 

It is worth clicking through to 'Lévy stable distributions' if you have an appetite for a bit of maths. Basically financial data series are fractal (by Mandlebrots definition) unless you disagree with his conclusions about the distribution of financial data series.

[Hlm 10]

jasont, maybe you should start a new thread called "Multi-Trend Strategies" and leave this one to discussing definitions and other mathematical concepts dealing with such a topic. Be sure to link this thread up front so as to make clear the difference between the two. Just a thought.

[firewalker 10]

  BlowFish said:

Basically financial data series are fractal (by Mandlebrots definition) unless you disagree with his conclusions about the distribution of financial data series.

 

Well, perhaps we should see what Mandelbrot himself had to say about the matter. This was most poignant:

 

"Financial market charts, however, are far from being self-similar."(*)

 

And that's all that matters. Lévy distributions, fat tails, chaos theory,... we could go on for days, but the assumption from some people is that markets are fractal in nature, ergo they exhibit the property of self-similarity.

 

This is not the case, and even Mandelbrot acknowledged that.

 

PS: yes I think it's best that we leave jason's thread alone and either move everything over or let him start a new one...

 

(*) "A Multifractal Walk down Wall Street", Benoit B. Mandelbrot, Scientific American, February 1999, pp. 70-73.

[BlowFish 10]

OK my last word on the subject then some practical observations of trading multiple time frames I do feel that FW is taking things out of context and feel that needs pointing out.

 

The title of the paper including the abstract/byline is

 

 

A Multifractal Walk down Wall Street

 

The geometry that describes the shape of

coastlines and the patterns of galaxies also elucidates how stock prices

soar and plummet.

 

by Benoit B. Mandelbrot

 

 

All Mandelbrot is saying is you can better model a price series with multiple exponents or multiple fractal dimensions. He is trying to model the 'shocks' the 'black swans' hence the reference to soar and plummet in the abstract.

 

From the paper : - "I claim that variations in financial prices can be accounted for by a model derived from my work in fractal geometry. Fractals—or their later elaboration, called multifractals—do not purport to predict the future with certainty. But they do create a more realistic picture of market risks."

 

Seems that you are taking things out of context to support your argument that markets are not fractal. (if that is indeed your argument). If people look at Mandelbrot's work they can make up their own minds. Of course most people won't bother which is the chief reason that there is confusion about terms, people talk about things they have no knowledge of as if they do. Anyway FW, I guess we can respectfully agree to differ though I think the idea of adding price distributions is pretty interesting and has a lot of relevance and practical application for profile type approaches.

 

OK on to the practical stuff

[OAC 10]

  CandleWhisperer said:

Markets are in fact self-similar. Take 5 charts from 5 different timeframes. Remove the time scale and the price scale. You will then be hard pressed to determine what timeframe is what.

 

I doubt anybody can dispute the above statement. Isn't the condition of non-distinguishability suffcient enough to be classified as being similar ? If not, what would be ? What is the definition of self-similarity anyway?

[firewalker 10]

  OAC said:

I doubt anybody can dispute the above statement. Isn't the condition of non-distinguishability suffcient enough to be classified as being similar ?

 

You are asking the wrong questions. When two objects, elements, patterns, charts or whatever exhibit a condition of "non-distinguishability" they are similar, because afaik those two words mean the same thing. And yes, you can show anybody a dozen charts and remove the timeframe and the will look alike.

 

However, self-similarity in the context of fractals is not the same thing which, again, might be the reason there seems to be a lot of discussion.

 

  OAC said:

If not, what would be ? What is the definition of self-similarity anyway?

 

If people would take the effort to find out, they would most likely not be arguing.

[firewalker 10]

  BlowFish said:

I do feel that FW is taking things out of context and feel that needs pointing out.

 

If you feel it's out of context, I'll just paste the whole alinea here:

 

"A fractal is a geometric shape that can be separated into parts, each of which is a reduced-scale version of the whole. In finance, this concept is not a rootless abstraction but a

theoretical reformulation

of a down-to-earth bit of market folklore—namely, that movements of a stock or currency all look alike when a market chart is enlarged or reduced so that it fits the same time and price scale. An observer then cannot tell which of the data concern prices that change from week to week, day to day or hour to hour. This quality defines the charts as fractal curves and makes available many powerful tools of mathematical and computer analysis.

 

A more specific technical term for the

resemblance

[does not equal exact copy]

between the parts and the whole is

self-affinity

. This property is

related

[but does not equal]

to the better-known concept of fractals called self similarity, in which every feature of a picture is reduced or blown up by the same ratio—a process familiar to anyone who has ever ordered a photographic enlargement.

Financial market charts, however, are far from being self-similar

."

 

I am only trying to make sure that no incorrect assumptions are taking place.

 

Mandelbrot said stock market charts appear self-similar when transformed via an appropriate anisotropic affine transformation for the level of detail being shown, a concept called self-affinity. Which is again, not the same as self-similarity, and in itself again debatable.